[骁龙峰会] 高通以ANF将主机级游戏体验带到移动端
高通在骁龙峰会上宣布,将其内部研发的Adreno Neural Fusion(ANF)技术用于移动设备,以带来主机级游戏体验。高通高级副总裁兼总经理Chris Patrick介绍了该技术。该消息与元器件供应链的关联在于,ANF作为高通自研的移动图形与AI融合技术,将随骁龙移动平台落地,影响移动SoC及相关元器件的需求与设计方向。
原文深度解读
文章报道高通在2026年骁龙峰会(Snapdragon Summit 2026,9月22日,毛伊岛)上,由高级副总裁兼移动手机业务总经理Chris Patrick宣布,通过自研的Adreno Neural Fusion(ANF)技术将主机级游戏体验带到移动设备。关键主体包括高通、骁龙8 Elite Extreme Gen 6平台、Adreno GPU、Adreno Matrix Core、18MB HPM以及Unreal Engine 5 MegaLights的Sanctuary演示。关键数字包括每帧超过3000个动态光源、每个光源独立实时光追阴影、光源数量为竞争对手两倍、光照复杂度为50倍、性能提升44%、能效提升40%、HPM带来12%能效改善。与电子元器件供应链的关系在于:ANF作为高通自研的移动图形与AI融合技术,将随骁龙移动平台落地,可能影响移动SoC、GPU相关IP、内存(LPDDR)及配套元器件的需求与设计方向。
- 高通高级副总裁兼移动手机业务总经理Chris Patrick在2026年9月22日于毛伊岛举行的骁龙峰会上表示,骁龙8 Elite Extreme Gen 6可实时运行Sanctuary演示,每帧超过3000个动态光源,且每个光源具有独立的实时光追阴影。
- Patrick称该演示的光源数量是竞争对手的两倍,光照复杂度是竞争对手的50倍,从而在移动端实现不妥协的主机级光照效果和图形。
- Sanctuary是基于Unreal Engine 5 MegaLights for Snapdragon的实时技术演示,其性能由ANF——高通新的基于AI的GPU技术——实现。
- ANF的硬件基础是新的Adreno Matrix Core:高通将Adreno GPU架构分为三个切片,并在每个块中加入专用AI矩阵计算核心,使图形管线所需的AI模型可直接在GPU内运行,从而减少外部计算数据的收发,降低延迟和浪费的功耗。
- 性能较上一代提升44%,能效提升最高40%;高通还在GPU中加入18MB High Performance Memory(HPM),减少对主LPDDR系统内存的访问,使能效提升12%。
- 软件方面,ANF集成原生AI图形渲染工作负载,结合超分辨率、帧生成和神经处理;ANF将3D渲染与AI图形整合到单一管线中,运行神经AI模型以高效计算复杂图形和光照效果,并称传统AI超分存在闪烁和重影等副作用。
供应链影响
- ANF随骁龙8 Elite Extreme Gen 6等移动平台落地,可能推动移动SoC中GPU与AI矩阵计算单元的集成度提升,进而影响相关芯片设计与代工需求。
- 由于ANF在GPU内加入18MB HPM以减少对主LPDDR系统内存的访问,移动终端的内存配置与带宽需求可能发生变化,具体取决于终端厂商的设计选择。
- ANF强调降低渲染成本与功耗,可能影响移动设备散热、电源管理及相关元器件的规格与用量,但实际影响取决于终端产品落地情况。
- 该技术依赖Unreal Engine 5 MegaLights等软件生态,可能带动移动游戏内容对高性能图形与AI算力的需求,从而间接影响相关元器件供应链的规格演进。
- 高通称ANF为从底层自研的专有技术而非现成方案,这可能强化其在移动GPU与AI图形领域的差异化,进而影响采用其平台的终端厂商的元器件选型方向。
系统解读,仅作为判断线索,不构成备货、出货、涨价或投资建议。
高通高级副总裁Chris Patrick在2026年骁龙峰会上发言。(照片:THE ELEC,毛伊岛)
高通正通过其内部研发的Adreno Neural Fusion(ANF)技术,将主机级游戏体验带到移动设备。
高通高级副总裁兼移动手机业务总经理Chris Patrick于9月22日在毛伊岛举行的2026年骁龙峰会上表示,骁龙8 Elite Extreme Gen 6可以实时运行“Sanctuary”演示,每帧超过3000个动态光源,并且每个光源都有独立的实时光追阴影。
“这是竞争对手光源数量的两倍,光照复杂度是竞争对手的50倍,从而在移动端实现不妥协的主机级光照效果和图形,”Patrick说。
Sanctuary是基于Unreal Engine 5 MegaLights for Snapdragon的实时技术演示。其性能由ANF实现,这是高通新的基于人工智能(AI)的图形处理单元(GPU)技术。
ANF旨在提供更高的图形质量,同时降低渲染成本和功耗。高通改造了其Adreno GPU的硬件架构,并结合了三项主要的AI游戏软件技术。
ANF较上一代性能提升44%,能效提升40%。(照片:THE ELEC,毛伊岛)
ANF的硬件基础是新的Adreno Matrix Core。高通将Adreno GPU架构分为三个切片,并在每个块中加入专用的AI矩阵计算核心。这使得图形管线所需的AI模型可以直接在GPU内运行。
在GPU内部处理AI工作负载减少了外部收发计算数据的需要,从而最大限度地降低延迟和浪费的功耗。性能较上一代提升44%,能效提升最高40%。
高通还在GPU中加入了18MB High Performance Memory(HPM)。减少对主LPDDR系统内存的访问使能效提升12%。
在软件方面,ANF集成了原生AI图形渲染工作负载。利用GPU内的AI加速,它结合了超分辨率、帧生成和神经处理。超分辨率使用AI将游戏图像提升到超过其原生分辨率,而帧生成使用AI提高每秒显示的帧数,以实现更流畅的游戏体验。
ANF将3D渲染和AI图形整合到单一管线中,并运行神经AI模型以高效计算复杂图形和光照效果。这实现了更复杂的基于AI的图形增强。传统AI超分存在闪烁和重影等副作用。
“ANF不是现成的解决方案,而是高通从底层设计的专有技术,”Patrick说。“通过将图形和AI紧密集成,它实现了新水平的智能游戏。”
Qualcomm Senior Vice President Chris Patrick speaks at Snapdragon Summit 2026. (Photo: THE ELEC, Maui)
Qualcomm is bringing console-class gaming to mobile devices with its internally developed Adreno Neural Fusion (ANF) technology.
Chris Patrick, Qualcomm senior vice president and general manager of mobile handsets, said at Snapdragon Summit 2026 in Maui on Sept. 22 that the Snapdragon 8 Elite Extreme Gen 6 can run the “Sanctuary” demo with more than 3,000 dynamic light sources per frame and independent ray-traced shadows for each light source in real time.
“That is twice the number of light sources and 50 times the lighting complexity of competitors, enabling uncompromised console-class lighting effects and graphics on mobile,” Patrick said.
Sanctuary is a real-time technology demo based on Unreal Engine 5 MegaLights for Snapdragon. The performance is enabled by ANF, Qualcomm’s new artificial intelligence (AI)-based graphics processing unit (GPU) technology.
ANF was designed to deliver higher graphics quality while reducing rendering costs and power consumption. Qualcomm revamped the hardware architecture of its Adreno GPU and combined three major AI gaming software technologies.
ANF improves performance by 44% and power efficiency by 40% from the previous generation. (Photo: THE ELEC, Maui)
The hardware foundation of ANF is the new Adreno Matrix Core. Qualcomm divided the Adreno GPU architecture into three slices and added dedicated AI matrix computing cores to each block. This allows AI models required during the graphics pipeline to run directly within the GPU.
Processing AI workloads inside the GPU reduces the need to send and receive compute data externally, minimizing latency and wasted power. Performance has improved 44% from the previous generation, while power efficiency has increased by up to 40%.
Qualcomm also added 18MB of High Performance Memory (HPM) to the GPU. Reducing access to the main LPDDR system memory improves power efficiency by 12%.
On the software side, ANF integrates native AI graphics rendering workloads. Using AI acceleration within the GPU, it combines super resolution, frame generation and neural processing. Super resolution uses AI to upscale game images beyond their native resolution, while frame generation uses AI to increase the number of frames displayed per second for smoother gameplay.
ANF integrates 3D rendering and AI graphics into a single pipeline and runs neural AI models to efficiently calculate complex graphics and lighting effects. This enables more sophisticated AI-based graphics enhancement. Conventional AI upscaling has suffered from side effects such as shimmering and ghosting.
“ANF is not an off-the-shelf solution, but proprietary technology that Qualcomm designed from the ground up,” Patrick said. “By tightly integrating graphics and AI, it enables a new level of intelligent gaming.”